Imec turns unstructured multimedia data into usable information.
Our data science researchers deal – amongst other things – with data quality, how to interpret and visualize data from different sources, and how to turn data into valuable information and (ultimately) knowledge.
Fusing semantic intelligence and artificial intelligence into a single focus, imec’s research teams across the various Flemish universities aim to solve five crucial data science research challenges.
A major challenge of current (big) data systems lies in the provision of high-quality information suited for downstream analysis purposes. Why? Because the data at hand are often irrelevant, redundant, noisy, inconsistent, incomplete or unstructured.
Imec research groups at UGent and KU Leuven are devising innovative cleansing, completion and feature extraction approaches to extract reliable information from raw data.
To make sense of big heterogeneous data, imec research groups at UAntwerpen, UGent and KU Leuven devise innovative mathematical/signal representations, statistical models, (deep) machine learning architectures and semantic reasoning techniques. They allow the analysis of distributed and non-distributed data in resource-limited and resource-abundant settings.
Data visualizations enable deep insights and new hypotheses. They can be aimed at a multitude of endpoints: 2D, autostereoscopic and light field displays and – the holy grail – holographic representations. They can be static or allow for rich user interaction such as brushing and filtering.
With the increasing size and complexity of datasets, the challenge of computational and perceptual scalability needs to be addressed. Imec research groups at UAntwerpen, KU Leuven and VUB alleviate this issue by distributed storage and computing as well as compression and visualization of the multimodal source data.
Big data analysis algorithms, statistical methods, querying engines and decision support heuristics all share one aspect: they are created to benefit us, humans. However, we lack the capacity to manually process all this data and its often highly technical analyses.
Imec targets the design of personalized and intelligent operational decision support agents who reason over the data for us. They provide us with understandable conclusions and recommendations.
Both are specialties of imec researchers at UGent and KU Leuven.
In the research domain of privacy and security, three long-term research challenges have been identified:
Imec’s security research aggregates more than 140 imec-KU Leuven researchers who’ve built up a top-notch track record of collaboration with academia and industry, both at a national and an international level.